R=19710004655 20180626T19:41:06+00:00Z
SPACEVEHICLE
DESIGNCRITERIA
(STRUCTURES)
FRACTURE CONTROL
OF
METALLICPRESSURE
VESSELS
MAY 1970
NASA experience has indicated a need for uniform criteria for the design of space
vehicles. Accordingly, criteria are being developed in the following areas of technology:
Environment
Structures
Guidance and Control
Chemical Propulsion
This monograph was prepared under the cognizance of the Langley Research Center.
The Task Manager was W. C. Thornton. The author was C. F. Tiffany of The Boeing
Company. A number of other individuals assisted in developing the material and
reviewing the drafts. In particular, the significant contributions made by C. P. Berry
and R.A. Rawe of McDonnell Douglas Corporation; D.W. Hoeppner of
LockheedCalifornia Company; R. L. Johnston of NASA Manned Spacecraft Center;
G. F. Kappelt of Bell Aerosystems Company; J. M. Krafft of the U. S. Naval Research
Laboratory; G. T. Smith of Lewis Research Center; H. G. McComb, Jr., of Langley
of Grumman Aircraft Engineering Corporation; J.C. Lewis of Jet Propulsion
Laboratory; G. T. Smith of Lewis Research Center; H. G. McComb, Jr. of Langley
Research Center; and C. D. Crockett of NASA George C. Marshall Space Flight Center
are hereby acknowledged.
May 1970
For sale by the National Technical Information Service, Springfield, Virginia 22151  P._ice $3.05
CONTENTS
I. INTRODUCTION .................... 1
3. CRITERIA ...................... 25
iii
4. RECOMMENDED
PRACTICES................ 28
B.I ThickWalledPressure
Vessel ......... 43
B.2 ThinWalledPressure
Vessel ......... 46
REFERENCES ......................... 51
SYMBOLS ........................... 55
NASASPACEVEHICLEDESIGNCRITERIA
MONOGRAPHSISSUEDTO DATE ................. 57
iv
FRACTURE CONTROL OF
METALLIC PRESSURE VESSELS
1. INTRODUCTION
Pressure vessels often contain small flaws or defects that are inherent in the materials
or introduced during a fabrication process. These defects can, in many cases, cause
severe reduction in the loadcarrying capability and the operational life of pressure
vessels. If the flaws are large in comparison to those causing failure at the
proofpressure stress levels, failure of the vessels will occur during initial pressurization.
If the initial flaws are small, the vessels may withstand several operational pressure
cycles and a number of hours of sustainedpressure loading before the flaws grow to a
size that will result in failure. From an economic standpoint, it is important to
minimize the possibility of failure of space vehicle pressure vessels during proof testing.
From the standpoint of economics and personnel safety, it is imperative to prevent
mission or operational failures.
During the past several years there have been costly prooftest failures directly
attributable to small, preexisting flaws. In one example, a large steel rocket motor case
failed at a stress less than 50 percent of the material yield strength. This failure
originated at a small internal flaw having a depth less than one fifth of the material
thickness. Other prooftest failures occurred in large propellant tanks and smaller
auxiliary tanks used in the Apollo program.
Other failures have occurred after proof testing during the preflight checkout and/or
storage of pressure vessels. One such failure occurred when a highpressure helium tank,
used in a defensive missile system, ruptured after 21 hours of sustained pressurization.
This failure originated at an inclusion in the parent metal. The initial flaw increased
approximately 50 percent in size during the time the tank was pressurized, and failure
resulted. Although this is an example of failure resulting from flaw growth under
sustained stress in a relatively inert environment, many more failures have occurred in
which the environment played the dominant role. A number of titanium pressure
vessels failed in N2 04 and methanol environments, and highstrength steel vessels failed
in water environments. In these cases, the initial flaw sizes were often small (i.e., less
than 10 percent of the size required to cause failure) and could not have been detected
by nondestructive inspection. However, with the vessels at pressure for a time, the
environmentinduced significant amountsof stable flaw growth and the vessels
eventuallyfailed.
The purposeof this monographis to presentcriteriaandrecommendpracticesthat aid
in the designof metallic pressurevesselsby minimizingthe occurrenceof prooftest
failures resultingfrom cracks and assuringagainstpreflight and flight failures.The
criteria andrecommended practicespermit widelatitude in the selectionof materials
andoperationalstresslevels,detail design,analysis,andtest to allowminimizationof
weight and/or cost asmay be dictatedby specificvehicleandmissionrequirements.
This monographis applicableto metallic pressurevesselswhosedesignis primarily
controlledby internalpressurerequirements.Thesevesselsincludehighpressure gas
bottles, solidpropellantmotor cases,and storable and cryogenicliquidpropellant
tanks both integral and removable.Criteria and recommendedpracticesfor the
design of pressurizedcabins, inflatable structures, and vesselsfabricated from
compositematerialswill bepresentedin othermonographsplannedfor this series.
2
The relatedproblemsof stresscorrosion,fatigue,anddiscontinuitiesin pressurevessel
designwill not be treatedherein,but will be coveredin other monographsnow in
preparation.
The problem of premature fracture of metallic structures is not new (e.g., the large
molasses tank failure in 1919, the methane storage tank failure in 1944, the 25percent
failure rate of Liberty ships during World War II, and the Polaris motor case failures
during the 1950's). Even today there is a general lack of specific guides in industry and
government manuals, specifications, and codes for the control of fracture of metallic
_k
pressure vessels. This results from the complexity and interdisciplinary nature of the
problem, the lengthy time required to develop and verify experimentally the technical
approaches, and the differing opinions on technical approach. The design of metallic
pressure vessels generally has been (and to some extent, is still) based on the following
principles:
1. The gross stress levels at the proof and operating conditions should be kept
below the yield strength of the material to prevent largescale deformations.
.
The fracture strength will be greater than the yield strength and equal to or
greater than the minimum guaranteed ultimate tensile strength of the
material.
,
Local yielding may occur around discontinuities, but the overall structural
integrity will be maintained by load relief and redistribution.
°
The factor of safety provides for uncertainties in stress analysis, fabrication,
and applied loads, and allows for possible degradation in strength with
service life.
o
Selection of factors of safety should be based primarily on experience, a
qualitative assessment of the uncertainties associated with a specific design,
and the reliability requirements.
= Sharpedged flaws or defects will not be allowed and, if any occur, they will
be detected by nondestructive inspection and subsequently repaired.
Althoughmany apparentlysuccessful pressurevessels
havebeendesignedaccordingto
the aboveprinciples,therehavebeenmany costly failuresat grossstresslevelswell
below the yield strength.In many of these cases,local yielding did not occur,
sharpedged flawsweremissedby inspection,andthe pastexperiences
usedin selection
of the factorsof safetywerenot applicable.
Theprimarylimitation of linearelastic
fracturemechanicsto dateis that at stresslevels
abovethe yield strengthof the material,fracturecannotbe describedby the critical
stressintensityparameter,Klc, and subcriticalflaw growthcannotbe describedasa
function of the cracktip stressintensityfactor, KI. From the standpoint of
application,this meansthat at stresslevelsabovethe yield strength,critical flaw size
andsubcriticalflawgrowthdatamustbeobtainedempiricallyovera rangeof flaw sizes
for the specificmaterialandthicknessof interest.Also,from the standpointof fracture
testing,it meansthat extremelythick test specimensare requiredto causefracture
prior to generalyieldingandthus obtain Kic valuesfor materialswith a highfracture
resistance(refs.2 and3). Anotherlimitation is the relativelysmallquantityof fracture
toughness andsubcriticalflawgrowthdatathat is generallyavailable.
Flaw types that often go undetected in metallic pressure vessels are the surface and
embedded flaws. The flaw size required to cause fracture at a given applied stress level
is called the critical size. If the vessel contains an initial flaw which exceeds the critical
size at the proofstress level, catastrophic failure can be expected during proof testing.
Failure during service operation will occur when the initial flaw is less than the critical
size at the proofstress level, but grows with service usage until it reaches the critical
size at the operating stress level. Pressure vessel leakage occurs when an initial flaw
grows through the thickness of the vessel wall prior to reaching critical size.
In elastic stress fields, the critical sizes for surface and internal flaws depend on the
planestrain critical stressintensity or fracture toughness values (Kic) of the vessel
materials, and the applied stress levels. If the critical flaw sizes are small with respect to
the wall thickness of the pressure vessel, the vessel is termed "thick walled." If the
critical sizes approach or exceed the wall thickness, the vessel is termed "thin walled."
The critical flaw sizes for surface flaws in uniformly stressed thickwalled vessels can be
calculated using the following expression:
1 (1)
(a/Q)cr 1.21ir
For small internal flaws the same expression can be used except the 1.21 coefficient is
decreased to unity.
Figure l shows the relationship between the flawshape parameter, Q, and the flaw
depthtolength ratio:figure 2 is a graphical representation of equation (1).
To predict critical flaw sizes (as well as failure modes and operational life) of
thinwalled pressure vessels, it is necessary to know the stress intensity for flaws that
become very deep with respect to the wall thickness. The stressintensity solution
shown in equation (1) for the semielliptical surface flaw was derived by Irwin (ref. 4)
and was found to be reasonably accurate for flaw depths up to about 50 percent of the
material thickness. At greater depths, the applied stress intensity is magnified by the
effect of the free surface near the flaw tip. This means that in thinwalled vessels, the
flawtip stress intensity can attain the critical value (i.e., the Kic value) at a flaw size
significantly smaller than that which would be predicted using equation (1).
0.5
k 2_1_1
r _ ____.,2a
'nterij
0.4
Surface flaw
o_" 1.0_
0.3 0.9_
0"8'X k
BQ)
6
v Y 0.5_
G/a, s = 0.6_
t Q = q_2 _ 0.212 (o/O )2
0.4_
0.2
"10
0.2 N"
i
u_
0
0.7 0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2
Flawshape parameter, Q
Oult
Eb ll\ 1
=
/ : _ lKlil 1" 1 ''f'_ 0 up (a/O) ` }_
• ; \_,c _2a_<,Co_Or,_
r_
<
I,, i I
"X.
' '%_  _,.= Operating stress
Gop
I ,% _L_ 
I I _" Max KI
Vlax I (a/Q) i
(a/Q)i I
I (a/Q)cr
I I
1 I ,
Flaw size, a/Q
....> •
Kobayashi and Smith developed approximate solutions for deep surface flaws that are
very long with respect to their depth (i.e.. small a/2c values) and for semicircular
surface flaws (i.e., a/2c = 0.5), respectively (refs. 5 and 6). Results of their solutions
are shown in terms of a stressintensity magnification factor, MK, versus a/t in figure 3.
Reference 7 shows an estimate made by NASA/MSC of how M K varies as a function of
a/2c between values of a/2c of 0 and 0.5. The M K. factor is applied to the original Irwin
equation to obtain the stress intensity for deep surface flaws. The magnification
reaches a maximum value of less than 10 percent for semicircular flaws, whereas there
is an increase of about 60 percent for flaws having smaller values of a/2c.
Experimental data obtained on several materials with varying flaw sizes and flaw shapes
appear to provide a fair degree of substantiation of the available approximate solutions
(ref. 8). An exact numerical solution for deep, semielliptical, surface flaws with varying
values of a/2c is under development, and additional experimental investigations are
being performed.
2.0
v
1.8
KI = 1.1 M K ,_O(a/Q) _
g
_ 1.6
?,
..r f
g J
I
E
/
_ 1.4 Kobayashi's solution ,,.,,
for small a/2c values /
_
I
1.2
(O/Oy s = 0.40)
Y
__fSomith's solution
r a/2c = 0.50 _
1.0
0.2 0.4 0.6 0.8 1.0
.1
:.. . .7 ¸ , [
a typical tank material and wall thickness is shown in figure 4. Also shown for
comparison is the critical flawsize curve for the same material in a thickwalled vessel.
The curve for the thinwalled vessel is characterized by a significant reduction in failing
stress at a given flaw size as compared to that for the thickwalled vessel. The life and
potential failure modes of these thinwalled vessels are schematically illustrated in
figure 5. The failure mode for thinwalled vessels can be complete fracture if the
critical flaw depth is less than the wall thickness at the operating stress level (figure
5A). Figure 5B illustrates the case where the critical flaw depth is greater than the wall
thickness at the operating stress level and the resulting failure mode is leakage.
From equation (1) it is apparent that to predict the critical sizes for surface and
internal flaws it is necessary to know the planestrain fracture toughness (Kic) values
for the vessel materials (i.e., parent metal, welds, etc.). In heavygage, highstrength
materials or in thingage materials that are relatively brittle, it is generally a
straightforward task to obtain Kic values from laboratory tests. Several types of test
specimens are used to measure Kicvalues. These include fatiguecracked bend
specimens, surfaceflawed specimens, crackline loaded specimens, centercracked and
edgecracked sheet specimens, and fatiguecracked round notchedbar specimens.
Testing requirements, limitations, advantages, and disadvantages of these various types
of test specimens are discussed in considerable detail in references 2 and 3.
70
I I 1 I
a/2c is small
MN
6O Kic = 37ksi i_. (1ks, i_. = 1.099 _v[m ")
AE
l\
z_:
LO
03
CO
5O
i\ \ Thinwalled tank
v = 1.1 M K _/_Oa _)
_ 40
/ / Thickwalled tank
e_
< thickness _ _ _"
2O
/Tank wall
10
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
"O
O
\ Max operating
>
 E
_j Fracture
<
jss E
ai.._
l_Wall
thickness
_, acr
_u/Leakag e
_r_
Max operatmg
O.
<
t str'_,,  EE=
_"Wall 7:
Thickness
To prevent failure, either the actual initial flaw sizes or the maximum possible initial
flaw sizes (or initial stressintensity factors) of pressure vessels must be known.
Nondestructive inspection is the only means of determining actual initial flaw sizes. A
successful proof test can provide a measure of the maximum possible initialtocritical
stressintensity ratio, and in turn allows the maximum possible initial flaw size to be
estimated.
The more common inspection techniques for inspection of aerospace pressure vessels
are radiographic, ultrasonic, penetrant, and magnetic particle. Other techniques
investigated for potential production usage include eddy current and infrared (ref. 13).
Several studies have been performed during the past several years to evaluate the
capabilities of these various techniques to detect the different types of flaws found in
pressure vessels (refs. 13 and 14). Results of these studies, combined with actual
pressurevessel inspection experience, lead to the following general conclusions:
. With the use of multiple inspection systems (e.g., Xray, ultrasound, and
penetrant), most surface and internal flaws encountered in pressure vessels
can be, and generally are, detected. However, it is unsafe to assume that all
potentially dangerous flaws will be found at all times (e.g., tight cracks are
particularly difficult to detect).
2. The lower limits of inspection detection capability (i.e., the largest initial
flaw sizes which can escape detection) cannot be confidently established.
10
190
180
170
E
Z
L_ Z_) _ 0
0"_
¢0 A \\ '___
co 160
±\
A \
\
150 _Oys
140
Grain Direction
O Longitudinal
130 
ATransverse
120
O 0.02 0.04 0.06
11
° The inspection procedures commonly used do not provide the precise
measure of initial flaw sizes (i.e., length and depth) necessary for use in a
fracture mechanics analysis.
For many years, it was normal practice to perform proofpressure tests on pressure
vessels; these tests, in effect, have served at least as one of the final inspections prior to
service usage of the vessels. However, prior to about 1960, very little was understood
regarding the determination of prooftest factors and prooftest procedures to
minimize potential damaging effects of the test, yet ensure adequate subsequent service
performance. During the past ten years, it has become apparent from the results of
fracture mechanics studies and aerospace pressurevessel experience that a properly
designed and successfully executed proofpressure test is probably the most reliable
nondestructive inspection technique available for insuring that there are no initial flaws
of sufficient size to cause failure under operating conditions.
From the standpoint of initial design, the minimum required prooftest factor for a
pressure vessel is ct = 1 + allowable KIi/KIc. The allowable value of Kii/Kic depends on
the required service life of the vessel and the subcritical flawgrowth characteristics of
the vessel materials and, ideally, should be a statistically meaningful value obtained
from laboratory test data.
Since the introduction of the prooftest concept, based on fracture mechanics, concern
has been expressed about possible damaging effects of the proof test; there has been
speculation that the test could cause the operational failure of a vessel that might have
12
performedsatisfactorilyhada proof test not beenperformed.Subcriticalflawgrowth
can,andoften does,occurin relativelyinert environments.Therefore,it is likely that
duringthe time requiredto performa proof test,initial flawsor defectsin the vessel
that areevidentcanincreasein sizeor possiblyflawswhich werenot evidentcouldbe
openedup. In fact, if the proof test is not properly designed(e.g.,if ct is < 1 +
allowable Kii/Kic , depressurization rates are too slow, or the test is conducted with an
aggressive test fluid), the flaw growth occurring during the test could be sufficient to
cause an operational failure.
During the past several years there have been numerous questions about the value of
the proof test with regard to the effects of applied stress levels and pressurevessel wall
thickness, selection of the test temperature, test fluids, pressurization and
depressurization rates, time at maximum pressure, multiple prooftest cycles, the need
for postproof inspection, and the need to simulate service loads other than internal
pressure. At present, there does not appear to be unanimity of opinion throughout
industry on the effects of these items. However, based on the premise that most
pressurevessel failures result from the existence and growth of flaws, several
observations and analyses can and have been made. These are summarized in the
following paragraphs.
To prevent general yielding during proof testing, pressurevessel membrane stresses are
normally limited to a value equal to or less than the yield strength of the material.
However, in practice, local stress levels often exceed the yield strength as a result of
design or manufacturing discontinuities and/or residual stresses. Also, in some cases
(e.g., cryoformed stainless steel vessels), the entire vessel may be purposely subjected
to stress levels well above the yield strength.
As shown in figure 2. when the applied stress approaches and exceeds the yield
strength of the material, the critical flawsize curve deviates from the theoretical curve
based on a constant Kic so that critical flaw sizes are smaller than those predicted by
linearelastic fracture mechanics. If the applied stresses in a pressure vessel at proof
pressure exceed the yield strength, and if the vessel passes the proof test, the maximum
possible Kii/Kic proven by the test is smaller than 1/a. The minimum operational life
of the vessel then should exceed the required life, which was used to determine a
originally. A potentially beneficial effect of high proofstress levels is that flaws may
tend to be blunted and, as a result, the subcritical flaw growth during operational use
of the vessel could be retarded. An apparent disadvantage is that at high proofstress
levels the critical flaw sizes may be very small compared to those that can normally be
detected; thus the prooftest failure rate may be quite high.
13
2.2.2.2 Effect of Wall Thickness
It has been shown by analysis that regardless of the pressurevessel wall thickness, the
required minimum prooftest factor a is 1 + allowable Kii/KIc. However, the value of
the proof test in providing assurance against service failure changes with decreasing wall
thickness and/or increasing fracture toughness, Kic, the same as occurs with the
predicted pressurevessel failure mode. This is discussed in more detail in reference 16
and illustrated in figure 7.
a _
The advantages of testing at a temperature where the value of Kic is lower than it is at
the operational temperature are as follows: (1) a lower prooftest factor can be used to
guarantee the same operational life as guaranteed by the corresponding higher
prooftest factor at the operational temperature, and (2) a larger operational life can be
assured by using the same prooftest factor as the one at operational temperature. The
disadvantage is the need to know accurately how Kic varies with temperature for all of
the materials in the vessel as well as the statistical variation in Kic for each material.
Also possible increased risk of prooftest failures is associated with the second case.
During the late 1950's it became apparent that the test fluid was often a major factor
contributing to the many prooftest failures that were being experienced. At that time
considerable emphasis was placed on the use of highstrength steel alloys in
solidpropellant motor cases, and it was common practice to perform the proof test
using water as the test fluid, One of the first systematic studies on the detrimental
effects of water on highstrength steel motor cases was performed by Shank et al. (ref.
17). In this study it was shown that by the mechanism of hydrogen cracking, the water
was promoting slow flaw growth that eventually resulted in failure of the motor cases.
With the use of oil as the prooftest fluid, the problem was overcome. Similar results
were obtained by researchers in other studies.
14
,i:i_ii_
..... • !:_::i i_ •
Wall Wall
/_KTH KTH
thickness.,_,_ th ickness,x, _
//KIc
/l<lc 'Kic \
Proof
aOop ....,. a'Oop a'Oop a'Oop
\
Oper__ !X ' %p Oop %p
°°P al i_ t Walt
Wall /
thickness / thickness
liar. I
I cr ial
LA
,, I I
I
Flaw depth, a Flaw depth, a Flaw depth, Flaw depth, a
Required
min. proof a= 1 _Allow. a = 1 : Allow. a = 1 : KTH/KIc a= 1 _ KTH/KIc
test factor
Kli/KI c Kli/KIc
The selection of the proper prooftest fluid is an important consideration for all alloys.
With precracked tensile specimens tested under sustained stress in the intended test
fluid, it is possible to obtain a measure of the adequacy of the fluid for use in the proof
test. (See sections 2.3 and 4.)
If the vessel is pressurized slowly, or if the proof pressure is sustained for a long period
of time, the probability of a prooftest failure is increased because of possible slow flaw
growth. However, after a successful test it can still be said that the maximum possible
Kii/Kic at the operating pressure is equal to 1/ct. On the other hand, if the vessel is
depressurized slowly so that the flaw that was just smaller than the critical size at the
proofstress level continues to grow, the maximum possible Kii/Kic after the test will
be greater than 1/ct_ In fact, it appears that if the rate of increase in stress intensity
caused by flaw growth is greater than the rate of decrease in stress intensity caused by
reduction in stress, the vessel could even fail during depressurization.
The amount of flaw growth that will occur during depressurization depends upon the
actual Kii/Kic ratio (or initial flaw size) at the start of depressurization, the
depressurization rate, and the flawgrowth characteristics of the vessel materials under
sustained stress in the prooftest fluid. If it is assumed that the Kii/Kic ratio
approaches unity (i.e., the vessel is just about to fail) at the start of depressurization,
and if sustainedstress flaw growthrate data for the material in the test fluid are
available, it is possible to determine the maximum possible Kii/Kic at the start of the
vessel's operational life as a function of depressurization time. This has been done for
some specific material and test fluid combinations in reference 7.
In general, it appears that very little can be gained by performing multiplecycle proof
tests. Even after the last cycle, all that can be said is that the maximum possible
Kii/Klc  1/ct, and that the cycles performed after the first cycle could have done some
16
needlessdamageto the vesselbecauseof cyclic flaw growth. However,special
circumstancesoccasionallydictatethe need,or makeit desirable,to conductmore
than one proof test. The majority of the vesselsusedin the Apollo programusea
singlecycle
proof test.
Current practice in industry regarding inspection after proof testing is divided, and
there have been arguments made both for and against this inspection. There is general
agreement that postprooftest nondestructive inspection can in some cases detect flaws
which were previously missed (perhaps because the flaws were too tight) and detect
flaw enlargement that may have occurred as a result of the proof test. Also, inspection
after proof test can potentially point to areas of the vessel requiring process or design
improvement. Considering this to be the case, the postproof inspection of at least the
initial vessels fabricated from a new design appears desirable.
However, the discovery of flaws following a proof test can create a dilemma concerning
the action required. If the flaws are repaired, another prooftest and postprooftest
inspection are generally required. This cycle could conceivably be repeated several
times before the vessel is (or appears to be) free of flaws. Furthermore, it is argued
(and many times correctly so) that the multiple repairs can be more detrimental than
the original flaws.
From the standpoint of fracture mechanics, there seems to be no particular need for
postprooftest inspection if the proof test is properly designed and successfully
executed. Any flaws that may be present after the test should not be of sufficient size
to cause operational failure of the pressure vessel.
In most proof tests of pressure vessels, internal pressure is the only applied load.
However, in some cases, vessels are critical for internal pressure combined with flight
loads, and it is not possible to represent the operational stress levels in the vessel by
internal pressure alone. In such cases, it generally appears desirable to include
provisions in the test setup to apply representative flight loads combined with
internal pressure. This has been done for some aerospace pressure vessels.
17
2.3 Subcritical Flaw Growth
Subcritical flaw growth can occur as a result of cyclic loading, sustainedstress loading,
and combined sustainedstress and cyclic loading. When the sustainedstress flaw
growth is environmentally induced, it is often termed stress corrosion; combined cyclic
and sustainedstress growth is called corrosion fatigue when environmentally induced.
Because of the potentially high rates of flaw growth, the problems of sustainedstress
and combined cyclic and sustainedstress flaw growth are particularly important in the
design of aerospace pressure vessels.
Data from fracture specimen tests can be used in a fracture mechanics analysis to
predict the number of cycles or the time the vessel must be under sustained pressure
for an initial flaw to grow to critical size. It has been shown (refs. 19 to 23) that for a
given environment and cyclic loading profile, the time or cycles to failure depends
primarily upon the magnitude of the initial stress intensity, Kii, as compared to the
critical stress intensity, Kic [i.e., cycles or time to failure = f(Kii/Kic)]. This is
particularly significant, because, as pointed out in the previous section, the proof test
provides a measure of the maximum possible Kii/Kic in the vessel.
During the past several years, cyclic and sustainedstress flawgrowth data have been
obtained for a large number of different pressure vessel materials in a wide variety of
environments. Although there are several methods of graphically presenting such data,
probably the simplest and most useful are plots of Kii/Kic versus cycles to failure and
Kii/Kic versus time to failure. Figure 8 shows typical Kii/Kic versus cycle data for
,.o i =, i _, 1 [
! ! , _! _ _. 1t]7 Bestfit
,east

I I I I I I I _t'_L"_ _ I I I/squarecurve
_x 0.6
_= H _/_c_nfidenTe II I I
o I (
, 1o ,oo ,ooo ,oooo
Cycles to fracture
18
6AI4V titanium at room temperature. Both the bestfit curve and the 96percent
probability and 99percent confidencelevel curve are shown. Figure 9 shows Kii/Kic
versus time data for 6A14V titanium in two different liquid, environments.
Experimental procedures used to obtain such cyclic and sustainedstress flawgrowth
data are described in several references including references 18 to 23.
Several different types of test specimens have been used to obtain subcritical
flawgrowth data. These include round notched bars, surfaceflawed specimens,
centercracked panels, singleedge notched specimens, crackline loaded specimens, and
notchedbend specimens. Of major interest to the pressurevessel designer is the growth
of flaws under planestrain conditions. Specimens containing throughcracks must be
relatively thick, for most materials, to develop planestrain conditions at the tip of the
crack. This requirement has restricted the use of such specimens for the thinwalled
pressurevessel life prediction problem. On the other hand, such specimens have the
advantage of permitting the observation and measurement of crack growth during the
course of the test. Acquisition of these data has not been limited to any one type of
specimen; however, the majority of the data on aerospace pressure vessel materials has
been obtained with the surfaceflawed specimen.
1.00
0.90
w, ;(JTH/KIc)_
v 0.80
o"
P 0.70
>
0.60
6i;;.n4vr_sT,_ )folging 4
E
0.50
P
0.40
0.30
b
O 0.20 water _
IKT./K,c>II
._m . • Fa,ore I I
0.10 No failure with flaw growth
E
Time, hr.
19
value of stress intensity, or Kii/Kic ratio, flaw growth has not been detected; above
this value, growth does occur and can result in fracture. This stress intensity has been
designated as KTH and is shown in figure 9.
The discovery of a unique KTH for a given material and environment is the key to the
design of safe pressure vessels subjected to sustained loading. While KTH can be 80
percent of Kic, or higher, in relatively inert environments, hostile media can reduce its
value to less than onehalf of Kic (fig. 9). In general, it has been found that KTH values
decrease with increasing yield strength in steel alloys (refs. 24 and 25). Also, there is
considerable evidence indicating that sustainedload flaw growth is most severe under
conditions of plane strain (ref. 26). Reference 27 shows that KTH values, determined
from tests of throughthethickness cracked specimens, increase with decrease in
specimen thickness.
Studies of flaw growth and stress intensity for materials in aggressive environments
(refs. 25 to 31) indicate an ever increasing flawgrowth rate with increasing stress
intensity; however, as shown in reference 7, the growth rate may be relatively constant
over an appreciable range of stress intensities. In tests for KTH, wide scatter is often
encountered in the data. Also encountered are abnormally short times to failure and
very marked dependence on environmental characteristics (media and temperature).
Even minor changes in the chemical composition of the environment can significantly
In chemically inert environments, the crack growth rate initially decreases with
increasing stress intensity. If the initialstress intensity is sufficiently low, the crack
may halt. At higher stress intensities, the crack growth rate passes through a minimum
value and then increases steadily until the crack becomes unstable. This flawgrowth
behavior is reported by Johnson (ref. 24) for AM 350 steel in a purified argon
environment.
This behavior is also noted in reference 20, where two threshold stress intensities were
defined for 5 A12.5 Sn (ELI) titanium and 2219T87 aluminum in the environments
of room air, liquid nitrogen, and liquid hydrogen. One threshold stress intensity was
defined as that value above which flaw growth to failure could be expected, and the
other as the value below which there is no flaw growth. In between these two
threshold stress intensities, small amounts of flaw growth can occur; however, the
growth apparently arrests after a short time at load.
From these remarks, it is apparent that the service conditions must be carefully
simulated when developing KTH data for pressure vessel design. Some examples of
experimentally determined KTH/KIc ratios are shown in table I.
20
TABLE I. TYPICAL THRESHOLD STRESSINTENSITY DATA FOR
VARIOUS MATERIAL/ENVIRONMENT COMBINATIONS
GTA welds:
Salt water >0.70 • "'33
18Ni (200) RT 200
steel spray
18Ni (250) RT 235 Salt water >0.70 33
steel spray
12Ni5Cr RT 170 Salt water >0.70 33
3 Mo steel spray
9Ni4Co RT 170 Salt water >0.70 33
2.5C steel spray
b 1 ksi = 6.895 MNIm 2. d No failure KTH, some growth observed at lower values (ref. 10).
21
Probably the most convincing evidence that the stressintensity factor, K, is the
controlling mechanical parameter in sustainedstress flaw growth are the strong
correlations obtained between various types of fracture test specimens and between
test specimens and actual pressure vessels.
Beachem and Brown (ref. 35) explored this consistency using three different test
specimen types:
Using 4340 steel in a dilute NaC1 solution, the same KTH value was obtained for all
three types of test specimens. The work of Smith, Piper, and Downey (ref. 28)
provides additional evidence. They used centercracked specimens to determine the
threshold stress intensity for crack initiation with end loading, and crack arrest with
wedgeforce loading. For Ti8AIIMoIV alloy in 3½ percent salt solution, the
threshold stress
intensity for crack initiation was 20 to 25 ksix/in"__
MN
(1 ksix/_. = 1.099_x/m) and for crack arrest 20 to 22 ksivqn. For endloaded
test specimens under constant load, both the stressintensity factor and net section
stress increase with increasing crack length; with wedgeforce loading, the net section
stresses increase whereas the stress intensity decreases with increasing crack length. The
excellent agreement between initiation and arrest values of KTH clearly shows that it is
the stressintensity parameter and not net section stress that is the controlling
parameter in sustainedstress crack growth. Correlations between sustainedstress flaw
growth in surfaceflawed fracture test specimens and pressure vessels subjected to
sustained pressurization are shown in references 10, 20, and 34.
22
2.3.2 Combined Cyclic and SustainedStress Flaw Growth
The use of Kii/Kic versus cycle data to predict the life of thickwalled pressure vessels
was first reported in the literature in reference 15. It indicated that if the maximum
possible Kii/Kic in the vessel were known (i.e., from a successful proof test), the
ordinate of a Kii/Kic versus cycles plot, such as that shown in figure 8, could be
entered at the appropriate value of Kii/Kic and the predicted minimum number of
cycles to fracture read from the abscissa. Experimental substantiation of this approach,
based on tests of actual preflawed pressure vessels, was subsequently presented in
references 10, 18, and 19. However, this approach was based on the assumption that
the pressure vessel was cycled at a speed comparable to that used in generating the test
specimen data or that cyclic speed was not important. In reference 2, it was
hypothesized that for values of initialstress intensity (Kii) below the sustainedstress,
thresholdstress intensity value (KTH), cyclic speed (or hold time at maximum load)
probably would not affect the cyclic growth rate of flaws; but for values of Kii above
KTH , it could have a significant effect. In other words, the minimum cyclic life was
limited by the number of cycles required to increase the value of Kii to the KTH value,
and above the KTH level, failure could occur in one additional cycle if the hold time
was sufficiently long: On a curve of Kii/Kic versus log cycles to fracture, this cyclic life
is represented by the difference between the number of cycles at the ordinates of
KIi/KI c and KTH/K I c
To date there are limited experimental data to substantiate this hypothesis. These data
were developed for 2219T87 aluminum and 5A12.5Sn(ELI) titanium in the relatively
inert environment of liquid nitrogen and are shown in reference 20. When materials are
subjected to more aggressive environments (i.e., those resulting in low KTH/KIc values)
there is considerable doubt regarding the general validity of the hypothesis. There are
some data on 8AIIMoIV titanium in a saltwater environment that indicate cyclic
frequency has no significant effect on flawgrowtla rate at stressintensity levels below
KTH. These data are shown in reference 37. On the other hand, recent investigations
by Barsom (ref. 38) and Wei (ref. 37) have shown that for some materialenvironment
combinations, both the environment and the cyclic frequency can affect the
flawgrowth rates at values of stressintensity below KTH. For example, Barsom has
shown that, for 12Ni steel in a salt water environment, cyclic growth rates of flaws are
higher than in a dry environment and progressively increase with decreasing cyclic
frequency (i.e., from 10 Hz to 0.1 Hz) at stressintensity (Kmax) levels less than KTH.
A complete explanation of this type of behavior has not been obtained; however, it is
apparent that additional research on environmentally enhanced fatigue growth (i.e.,
corrosion fatigue) is required.
23
If it is necessaryto usematerialshavinglowthreshold,stressintensityvalues(lessthan
70 to 80percentKIc) in the expectedoperatingenvironment,it appearsthat the
effect of environmentandcyclic frequencyon cyclic growthratesof flawsshouldbe
determinedandthe appropriateratesusedto estimatethelife of the pressurevessel.As
previouslymentioned,the minimum allowablecyclic life is limited to the numberof
cyclesrequiredto increasethevalueof the initial stressintensityKii to the KTH value.
The technique for using data on Kii/Kic versus cyclesto fracture to estimate
pressurevessel life also dependson pressurevessel
wall thickness.For thickwalled
vessels, the Kii/Kic curvescanbeuseddirectly,aspreviouslyindicated.For thinwalled
vessels,the task is somewhatmore complicated.Whenthe depth of a surfaceflaw
becomeslargewith respectto the wall thicknessof the vessel,the stressintensityis
higherthan that predictedby the originalIrwin surfaceflawequation(ref. 4), andasa
result, the subcriticalflawgrowthrateswill be higherand the total vessellife shorter
than that obtainedfrom Kii/Kic curvesof the type shownin figure8. (It shouldbe
noted that shallowsurfaceflawtest specimenswere usedin generatingthe basic
Kii/Kic data.) The increasein stressintensity for long surfaceflaws and for
semicircularsurfaceflaws, which becomedeep with respectto the vessel'swall
thickness,hasbeenapproximatedby KobayashiandSmith,respectively(Sec.2.1).As
indicatedin reference8 and shownin the examplein AppendixB, for thinwalled
vessels,it is necessaryto use flaw growthrate data and to account for the
stressintensitymagnification of deepflaws when making estimatesof vessel life.
Curves of flawgrowth rate can be obtained by differentiating the curves of Kii/Kic
versus cycle. For a given vessel design, the flaw growthrate curves can then be
arithmetically integrated using the Kobayashi approximation to account for the
increase in stress intensity as the flaw approaches the free surface of the pressurevessel
wall. A relatively simple procedure is shown in reference 8. Like thickwalled vessels
subjected to long hold times at maximum pressure, the cyclic life of thinwalled vessels
is the number of cycles required to increase the stress intensity from some known or
maximum possible initial value to the threshold value for sustained stress flaw growth.
In the analysis of thinwalled vessels, if it is found that the flaw gets very deep (i.e.,
approximately one plastic zone size from the back surface of the vessel wall) prior to
attaining the thresholdstress intensity, it appears wise to experimentally determine
cyclic flawgrowth rates with preflawed test specimens having the same thickness as the
actual vessel wall. The planestrain plastic zone size can be approximated by
(3)
24
Recentstudies(ref. 39) haveshownthat in this situationthe flawgrowthratesat a
given stressintensity
level may be higher than those predictedfrom the resultsof
shallowflaw,thickspecimen testdata.
3. CRITERIA
Metallic pressure vessels for space vehicles shall be designed to avoid service failure
caused by flaws and to ensure that the probability of catastrophic failure resulting
from flaws during proof tests is remote. The pressures, temperatures, environments,
and stresses from sources other than internal pressure to which the pressure vessels will
be exposed shall be accounted for. The materials selected for pressure vessels shall
possess appropriate fracture and flawgrowth characteristics; and, all material
properties or characteristics used in design and analysis shall be taken from reliable
sources of data or adequately substantiated by tests. Critical flaw sizes for stress levels
of interest shall be determined by analysis or test as appropriate. Where possible, the
maximum size of initial flaws permitted in pressure vessels shall be sufficient to have a
high probability of detection by nondestructive inspection but not sufficient to attain
the critical flaw size during the pressure vessel's service life. In addition, the permissible
initial flaw size shall be less than the critical flaw size at the proofpressure stress level.
The initial stressintensity ratio permitted in pressure vessels shall be selected to ensure
that the critical stressintensity ratio is not attained during the design life of the vessel.
Each pressure vessel shall be proof tested. The proofpressure level shall be selected to
demonstrate that the pressure vessel is free of flaws larger than the permissible initial
flaw size or that the actual initial stressintensity ratio is less than the permissible initial
stressintensity ratio. Account shall be taken of differences between the proof test and
service temperatures, and of the time required to pressurize and depressurize the vessel
during the proof test:
The maximum operating pressure shall be determined for each pressure vessel, and the
probability of exceeding this pressure during test (except proof test) and service usage
25
shall be sufficiently low to be consistentwith the overallvehicleflightworthiness
requirements.
The internalpressuretimetemperature
history for the vesselduringtest, storage,and
serviceuseshallbedetermined.
3.2 Materials
The fracture and subcritical flawgrowth characteristics of the pressure vessel materials
shall be determined for all critical environmental conditions.
Material properties used in the design of metallic pressure vessels shall be the "A"
values of MILHDBK5 for unflawed parent metal or obtained in the same manner as
those values.
26
_, _ ,/
' . , ,_ _ 14.
Material properties of weldments and repaired weldments shall be obtained by tests
based on the same procedure used in obtaining the "A" values of MILHDBK5 for
unflawed parent metal.
When the proof and maximumoperating stress levels are less than the tensile yield
strength of the pressurevessel material, the critical flaw sizes shall be calculated and
based on the appropriate stressintensity equations, the applied stress, and the
measured planestrain fracture toughness of the material.
When the applied stress (proof or operating) exceeds the tensile yield strength of the
material, the critical flaw sizes shall be empirically determined using test specimens
that contain flaws simulating those that could be encountered in the actual pressure
vessel.
The maximum permissible initial flaw size in metallic pressure vessels shall be the
largest flaw which cannot attain the critical flaw size within the required life span of
the vessel, and shall be smaller than the critical flaw at the proofstress level.
Pressurevessel joints having the permissible radial and/or angular mismatch and
containing the maximum permissible initial surfaceflaw size on the high
tensionstressed surface shall be capable of withstanding the proof stress without
failure.
The allowable initialtocritical stressintensity ratio for a metallic pressure vessel shall
be the largest value which cannot attain unity within the required life span of the
vessel.
The allowable initialtocritical stressintensity ratio shall be no higher than the value
obtained from an analysis of the subcritical flawgrowth tests of the pressurevessel
materials in the anticipated service environments.
The allowable initialtocritical stressintensity ratio for metallic pressure vessels subject
to shorttime pressurization shall be allowed to exceed the thresholdtocritical
stressintensity ratio only if it can be shown by test that the allowable ratio cannot
attain unity during the operational life of the vessel.
27
3.6 Proof Test
Each pressure vessel shall be subjected to a proof test. The prooftest factor shall be
equal to, or greater than, one divided by the allowable initialtocritical stressintensity
ratio.
When it has been shown by test that the pressurevessel materials exhibit a decreasing
fracture resistance with decreasing temperature, the proof test shall be conducted at a
temperature equal to, or less than, the lowest expected operating temperature.
The pressurization time and hold time at the proofpressure level shall be the minimum
practical, consistent with possible testsystem limitations. Emphasis shall be placed on
minimizing depressurization time.
Analytical and experimental verification that the probable service failure mode is
leakage rather than catastrophic fracture shall be required when assurance of safe
operational life cannot be provided by proof test.
4. RECOMMENDED PRACTICES
From the discussion in Section 2 it is apparent that to prevent prooftest failures, low
proofstress levels and materials having high fracturetoughness values should be used so
that the critical flaw sizes are large and hopefully exceed the thickness of the
pressurevessel wall. In this case the worst that could happen during proof testing is
that the vessel would leak and require repair. Also, it is apparent that to obtain
maximum assurance of safe operational performance it would be preferable to use large
prooftest factors, low operationalstress levels, and materials with low flawgrowth
rates under cyclic loads and high values of KTH in the expected service environment.
However, the use of high prooftest factors, low proofstress levels, low operatingstress
levels, and materials having very high fracturetoughness values (often associated with
low tensile strengths) generally leads to excessively high pressurevessel weight. With
the possible exception of some firststage launchvehicle tankage, these vessels are
generally not cost effective in terms of the delivery cost in dollarsperpound of
payload in orbit.
Tradeoffs can and should be made to arrive at an optimum design for a given pressure
vessel application. The interrelations between materials, the required service life of the
vessel, the required prooftest factor, the allowable flaw sizes, the probability of
prooftest failure, and the weight of the pressure vessel should be understood and
carefully assessed. These interrelations are illustrated in a simplified example in
Appendix A. Tradeoffs, however, must be made within the constraints provided by the
design criteria of the previous section.
28
4.1 Design Conditions
The value of maximum operating pressure used in the design of liquid propellant tanks
and gas bottles should equal the maximum nominaloperating pressure plus the upper
tolerance of the pressurelimiting device. This device should have a reliability consistent
with the overall vehicle flightworthiness requirements.
The predicted pressurevessel history should include pressures, times, temperatures, and
fluid and gaseous environments for all of the anticipated cycles, starting with the initial
proofpressure test and ending with the last servicepressure cycle. Also, it is important
to include pressurization rates, depressurization rates, and hold times. In those cases
where the life history of the vessel cannot be accurately predicted, a design life
envelope should be established and the appropriate operational limitations placed upon
the completed vessel.
Loads other than internal pressure, such as slosh, sonic, vibration, handling, and
transportation loads, should be determined in accordance with applicable NASA
monographs. Effort should be made to minimize high stresses resulting from flight and
ground loads by careful detailed design and by using antislosh, damping, and antishock
devices. Stresses resulting from external flight and ground loads should be determined
analytically and/or experimentally, and accounted for in the design of the pressure
vessel. Temperature gradients (and resulting thermal stresses) should be determined for
all critical ground and flight conditions. If the stresses are of sufficient magnitude to
affect the basic vessel design, an effort should be made to minimize or eliminate these
stresses using thermal insulation, controlled fill rates of cryogens, etc.
Wherever possible, the objective should be to eliminate residual stresses by stress relief
treatments. If this is not practical, residual stresses should be minimized by careful
design and controlled welding procedures.
A stress analysis should be performed for every vessel and include stresses resulting
from internal pressure, ground and flight loads, and thermal gradients. The analysis of
stresses resulting from internal pressure should include primary membrane stresses and
secondary bending and membrane stresses that result from design discontinuities and
allowable design deviations.
General yielding should be avoided during pressure testing except for those vessels that
are specifically designed to accommodate it (e.g., cryoformed stainlesssteel vessels). To
29
avoid generalyielding during proofpressuretesting, the minimum designultimate
factorof safety,(F.S.)MDU, shouldbeasfollows:
Where
The factors previously specified are minimum values for all metallic pressure vessels
used on both manned and unmanned vehicles. Uncertainties in loads, pressures, service
environments, and/or service requirements may make it necessary to use higher factors;
however, in no case should lower factors be used.
4.2 Materials
. The planestrain fracture toughness values (i.e., Kic values) for the parent
metal, weldments, and heataffected zones at the operating and prooftest
temperatures, and in the principal directions of loadings.
2. The threshold stressintensity (KTH) values for the parent metal, weldments,
and heataffected zones in simulated service environments.
d(_)
3. The cyclic flawgrowth data (curves of Kii/Kic versus cycles or _versus
K) for the parent metal, weldments, and heataffected zones.
The quantity of fracture test data obtained should be determined on the basis of the
impact a failure would have on the mission, schedules, and costs.
30
To comply with the criteria in this monograph,it is unnecessaryto limit the
determinationof fracture toughnessvaluesto any particulartype of test specimen.
However,it doesappearthat the curvesof predictedcritical flaw size(basedon the
measuredKic values) for the pressurevessel parent metal, weldments, and
heataffectedzonesshouldbeverifiedby datafroma seriesof surfaceflawed specimen
tests.The test specimens shouldbe the samethickness,processed in the samemanner
as the vessel,and eachshouldcontaina different sizeflaw. Procedures for specimen
fabricationandtest arediscussed in reference19.To eliminatethe effectsof inplane
bendingand specimenwidth, the testspecimen width shouldbe aboutfive times'the
surfaceflawlength(i.e.,the 2c dimension).
31
exhibit a lowthresholdstressintensity in the anticipatedserviceenvironmentshould
be avoided.If the materialhas a KTH value below about 70 percentof Kic, the
possibleuseof alternatematerialsshouldbeinvestigated.
. The critical flaw sizes associated with the increased proofstress level are
large (high probability of being detected prior to the test).
• _i!
•.... • 2. Sufficient experimental data are available to allow a reliable determination
of the biaxial improvement factor.
Because of the high probability of the occurrence of defects and the complexities in
stress fields introduced by design and manufacturing discontinuities, biaxial strength
elevation should not be used to establish allowable ultimate strengths of welded joints.
In cases where the effect of the biaxial stress field reduces the uniaxial tensile strength,
the amount of the reduction should be determined experimentally and used to
establish allowable strengths.
The concept of critical flaw sizes and the equations for determining these sizes for
surface flaws in thick and thinwalled vessels were introduced in section 2. These
equations apply, however, only when the gross stress levels of the pressure vessel are
below the yield strength of the pressurevessel material and when the stresses are
uniform through the thickness of the vessel wall. When this is the case (as in areas of a
vessel that are under membrane stress), it must be recognized that the accuracy of the
calculated critical flaw size depends directly on how accurately the material's fracture
toughness (Kic) and the applied stress levels are known. When calculating critical flaw
32
sizesfor theseareasof uniform elasticstress,the valueof Kic selectedfor designand
the maximumpossibleappliedstresslevel (i.e., that correspondingto the minimum
materialgage)shouldbeused.In addition,it is a conservative viewpointto assume that
the flaws are surface(or just subsurface)flaws andthat they arelong in relationto
their depthso that Q _ 1.0.Theresultingpredictedcriticalflaw sizeis thusdescribed
by the singledimension,a (i.e.,the depth).Whenthis depthis largewith respectto the
wall thickness(i.e., greaterthan about half the thickness),the effect of deepflaw
stressintensitymagnificationshouldbe accountedfor. The equationshownin figure3
attemptsto do this by the additionof the MK factor.A reasonable estimatefor MK is
the approximateKobayashisolution shownin figure 3. Whilerecentdata (ref. 39)
indicatethat its usecanresultin somewhatconservative answersfor the moreductile
materialsand perhapsslightly unconservative answersfor the brittle materials,it is
recommendedthat the figure 3 curvebe useduntil improvedsolutionsareobtained.
Sincethe equationshownin figure3 is not explicit in termsof the critical flaw size,
variouscritical depths(acr)shouldbeassumed for the longsurfaceflaw, the MK values
determinedfrom the Kobayashicurve,andthe failurestresses calculated.The curveof
u versusacr canthen be plotted.If the acr at the proof(or operating)stresslevelis
larger than the wall thickness,the expectedfailuremodefor the vesselat proof (or
operating)pressurewould beleakage.However,this canbepredictedwith confidence
only if there are no higherstressedareasin the vesselwherethe critical flaw depth
would be smaller,or if the valuecalculatedfor acr exceedsthe wall thicknessby a
significantamount.
33
discussed.For example,corner flaws, nearsurfaceinternal flaws, coplanarinternal
flaws, and sharptailedporosity may all be encountered.Again,for most of these
situations,reasonablyaccurateanalyticalestimatescanbe made(providingthe stress
field is elastic)usingvariousavailablestressintensity
solutions.Somesuchsolutionsare
includedin references 2, 6, 41, and42. Othersarecurrentlybeingdeveloped.
The two distinct areas of concern regarding initial flaw sizes are as follows:
Nondestructive inspection (i.e., Xray, ultrasonic, etc.) is the only means for
determining actual initial flaw sizes before the proof test (Sec. 2), consequently, such
inspections should be used to minimize the possibility of prooftest failure. The extent
of nondestructive inspection should be determined on an individual basis, taking into
consideration the consequences of a prooftest failure, the capabilities of the available
inspection techniques, and the sizes of initial flaws that must be detected (i.e., the
allowable initial flaw sizes).
The successful proof test provides a direct measure of the maximum possible
initialtocritical stressintensity ratio to predict the specific maximum possible initial
flaw sizes that may exist in the vessel after the proof test and before the service usage.
(Due to possible flaw growth during the proof test, the initial flaw sizes before and
after the proof test may not be the same). If the proof test is properly designed and
successfully executed, the maximum possible initial flaw sizes after the proof test are
equal to the predicted critical flaw sizes at the proofstress level. However, since the
proof test itself provides assurance against operational failure, the prevention of such
failure does not require the prediction of allowable initial flaw size.
Allowable initial flaw sizes should be determined for the following specific purposes:
34
Theserequire that the allowableinitial flaw sizesbe establishedfor all highstressed
areasof the vessel,includingthe parentmetal,weldments,andheataffected zones.
The allowable Kii/Kic ratio should be determined, using statistically meaningful curves
of subcritical flaw growth (i.e., KIi/Kic versus cycle and Kii/Kic versus time) and the
most severe service history anticipated for the vessel (Sec. 4.1).
The flawgrowth curves should take into account possible heattoheat variations in the
values of KTH and K Ic and the scatter in these values within a given heat. References
22 and 43 present discussions on the effects of data scatter and heattoheat variations.
35
4.6 Proof Test
Every pressure vessel should be proof tested to a stress level equal to or greater than
the maximum operating stress times ct (a = 1 + allowable Kii/KIc). If the vessel is
proof tested at a temperature other than the operating or service temperature, the
minimum prooftest factor, a. should be determined by equation (2) in Section 2.
In this case, it is important that the values of Kic are known for all areas of the vessel
and that it is known how they vary as a function of temperature. Also, it is important
to know the probable scatter in values of Kic at both the operating and prooftest
temperature. To ensure that the prooftest factor obtained will be adequate, the upper
statistical value of the Kic scatter band at the prooftest temperature and the lower
statistical value at the operating temperature should be used.
The proof test should be conducted with a test fluid that will neither induce general
corrosion pitting nor severe stresscorrosion cracking. The values of KTH for the vessel
materials should be obtained from sustainedstress fracture tests performed in the test
fluid at the prooftest temperature. If the values of KTH are low, either an alternate
fluid should be selected or, if this is not practical, methods of protecting the vessel or
inhibiting the action of the test fluid should be investigated.
Slow flaw growth during pressurization and elapsed time at proof pressure should be
minimized by rapid pressurization rates and short hold times. The pressurization time
should be the minimum possible, consistent with the capabilities of the test equipment.
A maximum hold time of about 15 seconds is considered to be reasonable.
Proof testing of metallic pressure vessels should be limited to a single pressure cycle
unless there are special circumstances indicating the need for additional cycles. Special
circumstances include the following cases:
,
The vessel was modified or repaired after the initial proof test and the
modified or repaired areas of the vessel need to be proof tested.
36
. It is desired to recertify the vessel for additional service usage after it has
been in service for a period of time.
.
From an economical standpoint, it is desired to test components (e.g.,
bulkheads) of the vessel prior to final assembly.
,
It has been shown by laboratory experiments on preflawed simulated parts
or specimens that a prior test at a higher temperature is advantageous to
minimuze the risk of failure at the design temperature.
Analytical and experimental verification that the probable failure mode is leakage
rather than complete fracture should be obtained in cases where assurance of
operational life is not provided by the proof test.
For those pressure vessels which are critical for internal pressure combined with flight
loads, it may not be possible to represent the operational stress levels in the vessel by
internal pressure alone. In such cases, the proof test should include provisions to
apply representative flight loads combined with internal pressure.
37
APPENDIX A
Figure A1 illustrates how the various factors affecting reliability and weight are
interrelated for pressure vessels designed to contain liquid hydrogen. In the upper
portion of the figure, the cyclic lives of two materials are shown as a function of the
inverse of the stressintensity ratio (Kii/KIc). The cyclic growth of initial defects or
flaws in a vessel is primarily a function of this ratio. Also, it can be shown that the
maximum possible Kii/Kic ratio in a pressure vessel after a successful proof test is
equal to 1 divided by the prooftest factor, a, or Kic/KIi = a. The solid lines are based
on the assumption of rapid pressure cycling where the sustainedstress flaw growth
above KTH is negligible. The dashed lines are based on the assumption that there are
longduration hold times at maximum pressure; and, consequently, the life is the
number of cycles required for the applied stress intensity to reach KTH.
In the center portion of the figure, constant flawsize lines are shown as a function of
the prooftest factor and the square of the ratio of the planestrain fracture toughness,
Kic , and the operational stress level aop. These curves were obtained as follows:
however:
Oproo f = a Oop
substituting:
1
(KIc 12 (A2)
\O_p/ = 3.8 a z (a/Q) i
With (a/Q) i held as a constant, the equation can be solved and plotted in terms of
(Kic/Oop) 2 vs a.
39
APPENDIX A
2.0 i
\ _1 I*tmax
1.8
1.2
Cycles to fracture
. _ _ ,
"_u 1.6 " i,/ _ k(>__(_j
1.0
0
I
_
A
I
0.2
!/,,
I J
0.4
f ,
'I
0.6 0.8
(Kic/Oop) 2
1.0
LL 1.5
>:
_r_ ,_. r
2.5 _ 4_/ _,.Co / P s
C3
5AI2.5 Sn (ELI) titanium
4,0 I ' I
Figure A1. Interrelated factors affecting the weight and reliability of thickwalled LH2 pressure vessels.
4O
APPENDIX A
The lower portion of the figure shows the relationship between the design ultimate
factor of safety (F.S.) and (Kic/Oop) 2 for the two materials, obtained from the
following relationship:
(Sop) \F.S._
°ult" 1/ t _ult. / (13)
Points of equal pressurevessel weight were computed for the aluminum and titanium,
and connected by dashed lines with the relative weight indicated.
It is doubtful if the titanium tank could successfully pass the proof test because of the
difficulty in detecting an initial flaw size as small as the critical flaw size at the proof
stress. On the other hand, this does not appear to be a problem with the aluminum
tank. The use of the conventional factor of safety of 2.5 seems to unduly penalize the
aluminum tank (i.e., causes it to be excessively heavy), and yet it is marginally
adequate for the titanium tank.
If an aluminum tank were designed with an ultimate factor of safety of about 1.75, its
weight would be equal to that of the titanium tank designed with an ultimate factor of
safety of 2.5, and the critical flaw size at the proofstress level (1.35 times Oop ) would
be about 0.09 in. This flaw is still about four times larger than that for the titanium
vessel and is sufficiently large to create some degree of confidence that all initial flaws,
41
APPENDIX A
From the foregoing example it is apparent that using standardized design factors does
not assure optimum (nor in some cases even adequate) designs. To preclude the
possibility of failure of hazardous vessels, high factors of safety have often been
specified. However, to save weight (caused by the high factors of safety) the designer
has been forced to use higher strength (and generally lower toughness) materials. As a
result, the risk of failure has often been increased rather than reduced.
While it can be argued that standardized factors of safety have been adequate for many
past applications, the designer must concern himself not with average behavior, but
with the exception which can result in failure. During recent years there have been
costly exceptions.
. • . J
42
APPENDIX B
ALLOWABLE STRESSINTENSITY
RATIO  ILLUSTRATIVE EXAMPLES
200 loading cycles with the maximum stress = 90 percent of Oop and
R = 0.1.
2. 4300 loading cycles with the maximum stress = aop and R  0.7.
3. 260 loading cycles with the maximum stress = 95 percent of aop and
R  0.4.
To design an adequate proof test for this vessel, it is necessary to determine the
maximum allowable Kii/Kic ratio and then to calculate the minimum prooftest factor.
The cyclic life curves for 6A14V titanium (STA) are reproduced in figure B2 for
R  0.1 and R  0.4, and R = 0.7 from reference 22. The change in Kii/Kic throughout
the life of the titanium tank is graphically illustrated in figure B2 and determined by
the following procedure.
Because the value of threshold stress intensity for sustainedstress flaw growth is 90
percent of Kic (table I), the allowable value of Kii/Kic at the beginning of the
longduration flight cycle at Oop is 0.90. This requirement is illustrated by point A in
figure B2.
The 60 loading cycles at Oop and R  0.1 change the Kii/Kic ratio from point A to
.,..,
point B in figure B2. Point B is 60 cycles to the right of point A, with the cycles being
measured along the abscissa of the plot of R  0.1. Hence, the allowable Kii/Kic ratio
at the beginning of the 60 cycles (point B in figures B1 and B2) is 0.84.
43
Relative applied stress or pressure, (_oo
Allowable stressintensity ratio, (K i i) max/Kic
o o o o
I"1
0 o ."
0 tn o bn
I I I I I I I I I I I I I I
[
[:J m,
::1 _ \S
"0
0 o
r, _ .oo
¢D t,Dto
3
D. O_ I I o
g 0
D. /
C_
C_
I
m__
m
C_
o_ :/ O
m
£3
i> 0"o o__ m
¢D m, z
4_
X
o 1_0'  ,1_,
(.0 ,_ _"
INI
 o_
2
Q _ =_
m. x "0
3
,<
3 m.
=_ r
o 0
&
c_ O
0
0
0
{)
m
m
L
m
13
0
II
__
o
/'J Q
II
" _LI °;
(1D
O'i
o_ g It /=,
"0 w oO !t?!._ Oil :
0
"0
 w I ,
0
0
APPENDIX B
The stress level is 5 percent lower at the end of the 260 cycles than at the beginning of
the 60 cycles, and since the flaw size is the same for both stress levels at that point
(point B in figure Bl), then the allowable value of Kii/Kic at the end of the 260
loading cycles is (0.95/1.00) times 0.84 = 0.798. This Kii/Kic ratio is given by point B
in figure B2 on the R = 0.4 curve.
The 260 loading cycles with the maximum stress = 0.95 Oop and R = 0.4 change the
Kii/Kic ratio from that given by point B to that given by point C in figure B2. Point C
is 260 cycles to the right of point B on the plot of R = 0.4. Hence, the allowable
Kii/Kic ratio at the beginning of the 260 cycles (point C in figures B1 and B2) is
0.74.
The stress level is 5 percent higher at the end of the 4300 cycles than at the beginning
of the 260 cycles and, by the same reasoning given above, the allowable value of
Kii/Kic at the end of 4300 cycles is (1/0.95) times 0.74 = 0.78. This Kii/Kic ratio is
given by point C in figure B2 on the R = 0.7 curve.
The 4300 loading cycles at Oop and R = 0.7 change the Kii/Kic ratio from point C to
point D in figure B2. Point D is 4300 cycles to the right of point C on the plot of
R = 0.7. Hence, the allowable Kii/Kic ratio at the beginning of the 4300 cycles (point
D in figs. B1 and B2) is 0.70.
The stress level is 10 percent lower at the end of the 200 cycles than at the beginning
of the 4300 cycles and therefore the allowable value of Kii/Kic at the end of the 200
cycles is (0.90/1.00) times 0.70 = 0.63. This Kii/Kic ratio is given by point D in figure
B2 on the R = 0.1 curve.
The 200 loading cycles with the maximum stress at 0.90 Oop and R = 0.1 change the
Kii/Kic ratio from that given by point D to that given by point E in figure B2. Hence,
the allowable Kii/Kic ratio at the beginning of the 200 cycles (point E in figs. B 1 and
B2) is 0.6. The operating stress is 10 percent higher than the stress at the beginning of
the 200 cycles so that the allowable value of Kii/Kic at the operating stress is (1.0/0.9)
times 0.6 = 0.667. This is shown by the asterisk in figure B2.
45
APPENDIX B
Thus, for the pressure vessel subjected to the anticipated service history given, the
maximum allowable Kii/Kic ratio at the end of the proof cycle is 0.667 and the
minimum required prooftest factor is a = 1/0.667 = 1.5. This indirectly imposes a
restriction on the maximum allowable operating stress because the proof stress should
not exceed the yield strength of the material. Hence, the maximum allowable operating
stress is 0.667 times Oy s.
In the thinwalled tank, the flaw depth becomes deep with respect to the wall thickness
of the tank before reaching the critical size. Hence, the stressintensity factor must be
corrected for the a/t ratio according to figure 3. Suppose the thickness of the tank wall
is 0.022 in. (1 in. = 0.0254 m) and the maximum design operating stress, aop is
84.4 ksi (1 ksi = 6.895 MN/m2). Under the specified environmental conditions, the
material of this gage has a minimum fracture toughness of 37 ksi_/in.
(1 ksi x/_. = 1.099 M_N _m) and a threshold stress intensity of 80 percent of Kic. The
111
plot of flawgrowtl] rate versus Kii/Kic for the material is shown in figure B4 for
o = 105 ksi (1 ksi = 6.895 MN/m2). The effect of the stress level on the growth rate is
indicated by the equation on the plot. Taking this effect into consideration, the curve
is arithmetically integrated, according to the method outlined in reference 8, for three
stress levels. These integrated plots (flaw depth versus cycles to fracture) are shown in
figure B5. In the calculations, it was assumed that the value of Q is unity (i.e., the
flaws are relatively long with respect to their depth).
46
APPENDIX B
1.50op _
Proof cycle
.=
==
t_
0.50"op 
i
cr
0
Cycles
1.0
L) _.m
J
v 0.8
I
Z J
o" 0.6
I
I "_0" = 105 ksi (1 ksi = 6.895 MN/m 2)
.E
0.4 I I
/lo5_2
"d(a/e)/dN]o 1 =[d(a/e)/dN]o" = 105 x \ (/1 ]
',_ 0.2
L. 6AI4V titanium STA forging
.o MN
9
Kic = 37 ksi_/_. (1 ksiv/_. = 1.099 
2
m
0.1
I II I
10 100
d(a/Q)/dN,/_ in./cycle
47
i')
APPENDIX B
Because the threshold stress intensity is 0.80 Kic, the allowable value of Kli/Kic at the
beginning of the longduration flight cycle is 0.80. This requirement is illustrated by
The tankwall stress increases by 11 percent at the end of 20 loading cycles with the
maximum stress = 0.89 aop; however, the flaw size remains the same during the stress
increase• This is shown by point A on the plot of 0.89 aop in figure B5.
The 20 loading cycles with the maximum stress = 0.89 Oop changes the flaw depth (a)
from point A to point B on the plot of 0.89 aop in figure B5. Point B is 20 cycles to
the right of point A with the cycles being measured along the abscissa of the plot.
The stress decreases by 11 percent at the end of 9 cycles with the maximum
stress = Oop. This is shown by point B on the plot of Oop in figure B5.
0.017
_g
LO
_q
0
d 0.016
. 0.015
0.014
LL
0.013
0.012
1 10 100 1000
Cycles to fracture
48
APPENDIXB
The 9 loading cycles with the maximum stress = Oop changes flaw depth (a) from point
B to point C on the plot of Oop. Point C is 9 cycles to the right of point B.
The tankwall stress increases by 5 percent at the end of 20 loading cycles with the
maximum stress = 0.95 aop. This is shown by point C on the plot of 0.95 Oop in figure
B5.
The 20 loading cycles at 0.95 Oop changes flaw depth (a) from point C to point D on
the plot of 0.95 Oop. Point D is 20 cycles to the right of point C. The value of the flaw
depth at point D is 0.01356 in. (1 in. = 0.0254 m).
The maximum allowable value of Kii/Kic at the end of the prooftest cycle then is
given by
Oop = 84.4 ksi (1 ksi= 6.895 MN/m2), a/t = 0.013561'0.022 = 0.615, and MK from
figure 3 is 1.25.
Hence, the maximum allowable Kii/Kic ratio is 0.647, and the proof factor is
a= 1/0.647 = 1.55.
49
i ¸¸¸ _ S _
I
r
L_
• i I kl ¸ I
REFERENCES
.
Wessel, E. T.; Clark, W. G.; and Wilson, W. K.: Engineering Methods for the
Design and Selection of Materials Against Fracture. U.S. Army Tank and
Automotive Center Report, Contract No. DA30069AMC602(T), 1966.
.
Tiffany, C. F.; and Masters, J.N.: Applied Fracture Mechanics. Fracture
Toughness Testing and its Applications. ASTM Spec. Tech. Publication No. 381,
1965, pp. 249277.
.
Brown, W. F., Jr.; and Srawley, J. E.: Plane Strain Crack Toughness Testing of
High Strength Metallic Materials. ASTM Spec. Tech. Publication No. 410, 1966.
°
Irwin, G. R.: Crack Extension Force for a PartThrough Crack in a Plate. Trans.
ASME, J. Appl. Mech., Series E, vol. 29, no. 4, Dec. 1962, pp. 651654.
.
Kobayashi, A. S.: On the Magnification Factors of Deep Surface Flaws.
Structural Development Research Memorandum No. 16, The Boeing Co., Dec.
1965.
.
Smith, F. W.: Stress Intensity Factor for a SemiElliptical Flaw. Structural
Development Research Memorandum No. 17, The Boeing Co., Aug. 1966.
.
Tiffany, C. F.; Masters, J. N.; and Bixler, W. D.: Flaw Growth of 6A14V
Titanium in a Freon T.F. Environment. NASA CR99632, 1969.
,
Tiffany, C. F.; Masters, J. N.; and Pall, F. A.: Some Fracture Considerations in
the Design and Analysis of Spacecraft Pressure Vessels. Paper presented at the
ASM National Metals Congress (Chicago), Oct. 1966.
,
Anon.: ASTM Recommended Practice for PlaneStrain Fracture Toughness
Testing of High Strength Metallic Materials Using a Fatigue Cracked Bend
Specimen. ASTM Standards, Part 31, May 1969, pp. 10991114.
11. Hall, L. R.; and Tiffany, C. F.: Fracture and Flaw Growth Investigation for
2014T6 Aluminum Weldments Used in Saturn II LH2 Tanks. Rept. D515737,
The Boeing Co., Nov. 1967.
51
12. Smith, A. B.: MissileMotor Cases.MetalsEngineeringQuarterly,vol. 3, no. 4,
Nov. 1963,pp.5563.
17. Shank,M. E.; Spaeth,C. E.; Cook,V. W.;andCoyne,J. E.: Solid Fuel Rocket
Chambersfor Operationat 240,000psiandAbove.MetalProgress, vol. 76,no. 5,
Nov. 1959,pp. 7481(part I), andvol. 76, no. 6, Dec.1959,pp. 8492(part II).
21. Tiffany, C. F.; and Masters, J. N.: Investigation of the Flaw Growth
Characteristicsof 6A14VTitanium Usedin Apollo SpacecraftPressure
Vessels.
NASACR65586,1967.
52
24. Johnson,H. H.; andParis,P. C.: SubcriticalFlaw Growth.Engineering
Fracture
Mechanics,
vol. 1.no. 1,June1968,pp. 345.
25. Peterson,M. H.; Brown, B. F.; Newbegin,R. L.; and Grover, R. E.: Stress
CorrosionCracking of High StrengthSteel and Titanium Alloys in Chloride
Solutionsat Ambient Temperature.Corrosion,vol. 23, no. 5, May 1969. pp.
142148.
26. Brown, B. F.: A New Stress Corrosion Cracking Test Procedure for High Strength
Alloys. Paper presented at the ASTM Arsenal Meeting at Purdue University
(Lafayette, Ind.), June 1318, 1965.
27. Piper, D. E.; Smith, S. H.; and Carter, R. V.: Corrosion Fatigue and Stress
Corrosion Cracking in Aqueous Environments. Met als Engineering Quarterly, vol.
8, Aug. 1968, pp. 5063.
28. Smith, H. R.; Piper, D. E.; and Downey, F. K.: A Study of Stress Corrosion
Cracking by Wedge Force Loading. Engineering Fracture Mechanics, vol. 1, no. 1,
June 1968, pp. 123128.
29. Johnson, H. H.; and Willner, A. M.: Moisture and Stable Crack Growth in a High
Strength Steel. Applied Materials Research, vol. 4, no. 1, Jan. 1965, pp. 3440.
30. Steigerwald, E. A.; and Benjamin, W. D.: Stress Corrosion Cracking Mechanisms
in Martinistic High Strength Steels. Third Quarterly Progress Report, Contract AF
33(615)3651, Air Force Materials Laboratory, 1967.
31. Irwin, C. R.: Moisture Assisted Slow Crack Extension in Glass Plate.
Memorandum Report 1678, Naval Research Laboratory, 1966.
33. Tiffany, C. F.; Masters, J. N.; and Regan, R. E.: Large Motor Case Technology
Evaluation. Rept. AFMLTR67190, Air Force Materials Laboratory, Aug. 1967.
34. Lorenz, P. M.: Fracture Toughness and Sustained Flaw Growth Characteristics of
Inconel 718 in the Environment of Pressurized Gaseous Hydrogen. Rept.
D21144041, The Boeing Co., Oct. 1968.
35. Beachem, C. D.; and Brown, B. F.: A Comparison of Three Specimens for
Evaluating the Susceptibility of High Strength Steel to Stress Corrosion Cracking.
Internal Report, U.S. Naval Research Laboratory, 1967.
53
36. Haese,W. P." Investigationof Fracture of 6AI4V Titanium in N204. Rept.
D2240571, The Boeing Co., Dec. 1965.
39. Masters, J. N.; Haese, W. P.; and Finger, R. W.: Investigation of Deep Flaws in
Thin Walled Tanks. NASA CR72606, 1969.
40. Shah, R. C.; and Kobayashi, A. S.: Stress Intensity Factor for an Elliptical Crack
Under Arbitrary Normal Loading. Paper presented at the Second National
Symposium on Fracture Mechanics at Lehigh University (Bethlehem, Pa.), June
1719, 1968.
41. Kobayashi, A. S.; Ziv, M.; and Hall, L. R.: Approximate Stress Intensity Factor
for an Embedded Elliptical Crack Near Two Parallel Free Surfaces. Int. J. of
Fracture Mechanics, vol. 1, no. 2, June 1965, pp. 8195.
42. Shah, R. C.; and Kobayashi, A. S.: On the Parabolic Crack in an Elastic Solid.
Engr. Fracture Mech., vol. 1, no. 2, Aug. 1968, pp. 309325.
43. Lorenz, P. M.: Compatibility of Tankage Materials with Liquid Propellants. Rept.
AFMLTR6999, Air Force Materials Laboratory, May 1969.
54
. • . , • .:v
SYMBOLS
N number of cycles
T time, hr
a prooftest factor
0 angle of integration
SUBSCRIPTS
cr at critical conditions
i at initial condition
op operational
55
 . , '. .
L••
r
NASA SPACE VEHICLE DESIGN CRITERIA
MONOGRAPHS ISSUED TO DATE
57
SP8026 (Guidance SpacecraftStarTrackers,July 1970
andControl)
SP8027 (Guidance SpacecraftRadiationTorques,October1969
andControl)
SP8028 (Guidance Entry VehicleControl,November1969
andControl)
SP8029 (Structures) AerodynamicandRocketExhaust HeatingDuring
LaunchandAscent,May 1969
SP8031 (Structures) SloshSuppression,
May 1969
SP8032 (Structures) Buckling of ThinWalledDoubly Curved Shells,
August1969
SP8033 (Guidance SpacecraftEarthHorizon Sensors,
December1969
andControl)
SP8034 (Guidance Spacecraft
MassExpulsionTorques,December1969
andControl)
SP8035 (Structures) Wind Loads During Ascent, June 1970
SP8036 (Guidance Effects of Structural Flexibility on Launch Vehicle
andControl) Control Systems, February 1970
SP8046 (Structures) Landing Impact Attenuation for NonsurfacePlaning
Landers, April 1970
58 NASALangley, 1970 32
r •